Two experiments were conducted with 26 Holstein and Brown Swiss cows, the first (E-1) during the dry season (February to April) and the second (E-2) during the rainy season (July to September). Treatment L, which involved 3 hours daily of grazing in protein banks of leucaena and associated grasses, was compared with a control (T) of grazing in grass herbage only. The paddocks of grass alone received no fertilizer. Vegetation was scarce during the dry season, but abundant during the rainy season. Individual supplementation with concentrate feed (16% crude protein, CP) was according to production, but the maximum daily allowance was fixed at 7 kg. The experimental design was a cross-over (single reversal of treatments) with 21-day comparison periods. The three herbages— grasses alone, grasses in association with leucaena, and the leucaena— showed the following mean percentage chemical compositions in E-1 one B-2, respectively: CP, 5.9, 6.9 and 20.3; 7.3, 6.4 and 23.7; acid detergent fiber, 34.4, 34.3 and 28.2; 45.8, 44.3 and 27.6. Samples of rumen fluid of L and T cows had 179 vs. 148 mg/l of ammonia in E-1 and 184 vs. 159 mg/l in E-2. Total and proportional concentrations of volatile fatty acids and pH of the rumen liquor revealed no consistent effect of the treatments. Lactational responses observed with L and T were: daily milk, 1 5.1 vs. 14.5 kg (P = 0.05) in E-l and 15.6 vs. 15.6 kg in E-2; fat percentage, 3.1 vs. 3.2 in both experiments; protein percentage, 3.7 vs. 3.6 and 3.2 vs. 3.3 in E-1 and E-2. Daily consumption of concentrates was 5.6 vs. 5.7 kg and 6.5 vs. 6.4 kg; and liveweight 542 vs. 540 kg and 506 vs. 497 kg for L and T in the two successive experiments. Presumably the positive effect of the protein bank on milk production during the dry season was due to increased herbage intake.